Packing sorted by an attribute?

[probiner 83]

Given weighted (colored red:x, green:y) polygon islands if I use the UV Layout SOP, the colors will have somewhat random distribution:

 

Now, what I'm wondering is if there's an already established way to have the packing respect an attribute (vector2 or 2 floats) to sort the uv islands according to one or both values in the respective axis. Here's a mock-up:


Even if I don't use UVs and do some simulation, it's not sure how I would go about to have the, let's say cylinders, fill a space and respect sorting.

Scene: PRB_WeightedPacking.hiplc

Cheers
prb

Edited December 18, 2021 by probiner

[konstantin magnus 1,163]

Matt published a perhaps suitable packing method called "Pack fattest to thinnest boxes":
https://www.tokeru.com/cgwiki/index.php?title=ForLoops

[vinyvince 117]

Yes, UV layout is a complex operation to build from scratch , if you are interested, you could learn more about the theory there by looking for this  paper

Least squares conformal maps for automatic texture atlas generation" by levy et al, https://members.loria.fr/Bruno.Levy/papers/LSCM_SIGGRAPH_2002.pdf

 

I agree with Konstantin Solution, looks like a solution to me. Matt delete a random piece, but you could establish rule and  as copy on the UV 0.1 space

 

 

 

[probiner 83]

On 12/24/2021 at 10:03 AM, konstantin magnus said:

Matt published a perhaps suitable packing method called "Pack fattest to thinnest boxes":
https://www.tokeru.com/cgwiki/index.php?title=ForLoops

Thanks, but can't use this dart throwing method in my case, since I have a fixed set of islands to pack, so I can't just delete part of them for not fitting. It's a useful method, just not for my case.

 

On 12/26/2021 at 2:41 PM, vinyvince said:

Yes, UV layout is a complex operation to build from scratch , if you are interested, you could learn more about the theory there by looking for this  paper

Least squares conformal maps for automatic texture atlas generation" by levy et al, https://members.loria.fr/Bruno.Levy/papers/LSCM_SIGGRAPH_2002.pdf

 

I agree with Konstantin Solution, looks like a solution to me. Matt delete a random piece, but you could establish rule and  as copy on the UV 0.1 space

 

 

 

As mentioned above, deletion method won't do for me. Thanks for the link about LSCM, I'll see if it inspires me to get something functional going. But let's see if some ready made approach surfaces

Cheers
prb

Edited December 28, 2021 by probiner

[Keshaw singh 0]

you can find this solution in mops system